Re: dataset and Dataset

Hi Raphaël,

On 6 Apr 2013, at 18:13, Raphaël Troncy <raphael.troncy@eurecom.fr> wrote:
>> Multiple popular vocabularies that use this practice have been listed
>> in the thread.
> 
> I have seen those. I still think this is rather anecdotal in comparison to the very large set of properties that has a hasXXX OR a isXXXBy  that one can found in the wild. Doing a quick search on LOV confirms me this.

Sure, there are lots of vocabularies that contain some properties that are named hasXxx or isXxxOf. That's undisputed and besides the point.

The question is whether the hasXxx/isXxxOf convention is so universally accepted that DCAT would do wrong by not following it. This is clearly not the case. DC, FOAF and SKOS, and the RDF and RDFS vocabularies themselves don't use a hasXxx/isXxxOf convention.

>> That's a convention coming from Description Logics that I haven't
>> seen in any other modelling school.
> 
> Don't count on me to fuel any religious wars. I don't care where does it come from and I don't think it matters. I'm not myself feeling part of the DL school (as I suspect many) and I still model my vocabularies this way.

That is fine, but please accept that there are lots of other people who *don't* model their vocabularies this way.

>> I'm with TimBL on this one:
> 
> I fully agree with this too, but this is really an orthogonal issue. I'm not saying that one should model both a property and its inverse in the same ontology.

Okay, I misunderstood.

> I'm saying that when the ontology engineer has chosen the direction of the property he wants to model, then the label should convey this direction. Otherwise, the person who wants to re-use the property has to look at the domain and range, and not get confused of what is the domain and the range [as most of the people are, this is the same well-known problem that turning left/right with your car and there are many studies that show this :-)]. The hasXXX OR isXXXBy pattern fulfills this use case.

By convention, a property xxx (where xxx is a noun) means "has xxx". This is blindingly obvious outside of the SemWeb world, when we're talking about fields on an object or columns of a table or attributes of an XML element. With my RDF hat off and my data modeller hat on, I don't really see why we should name things differently just because we use RDF.

It is true that in *some* cases it's very easy to get confused about the direction of RDF properties. Especially with properties where domain and range are the same, such as "part". I guess that's why DC, which usually uses simple nouns for relationships (dc:creator, not dc:hasCreator), uses dct:partOf and dct:versionOf when relating two resources of the same type.

I agree with Dave that in DCAT there's not much of a risk of this confusion.

Best,
Richard

Received on Saturday, 6 April 2013 18:10:37 UTC